Machine-Learning Prediction of Weak Buses and Optimal SVC Placement for Voltage Stability Improvement on the Transmission Network with Special Emphasis on Nigeria

Abstract

Voltage instability remains one of the most persistent threats to secure transmission network operation, particularly in developing power systems where reactive power reserves are thin and infrastructure investment lags behind load growth. This review examines two research streams that have largely progressed in parallel: machine-learning-based identification of weak buses and metaheuristic-driven optimal placement of static VAR compensators (SVC) for voltage stability enhancement, using the Nigerian 330 kV transmission network as an illustrative case of a stressed developing-country grid. The review traces the evolution of voltage stability indices, from the foundational L-index through the fast voltage stability index (FVSI), and evaluates how supervised learning models, including random forest, gradient boosting and neural architectures, have been applied to accelerate weak-bus ranking beyond the limits of conventional load-flow-based screening. It further synthesises evidence on genetic, particle swarm and gradient-based metaheuristics for SVC and related flexible alternating current transmission system (FACTS) placement, and situates these methods within the technical and institutional constraints of the Nigerian grid, where chronic under-compensation and recurrent system stress create acute demand for scalable, data-driven planning tools. The review finds a persistent disconnection between weak-bus prediction research, dominated by small test-system validation, and SVC placement research, which rarely incorporates learned stability predictions as a planning input. It concludes that hybrid frameworks coupling machine-learning-based weak-bus screening with metaheuristic compensator placement offer the most promising route towards resilient, cost-effective voltage stability management on constrained transmission networks, while cautioning that data scarcity, model interpretability and the absence of frameworks validated on the Nigerian network remain significant barriers to practical deployment.